Various small cleanups and bugfixes

[jcupitt]

A few smallish things I found while getting the openslide connnector working.

[jcupitt]

Sorry, the history on this is a bit of a mess.

Summary:

1. I needed to be able to get tags from keywords, so I added another hash. I named the keyword <-> tag functions as :c:func:dcm_dict_tag_from_keyword() and :c:func:dcm_dict_keyword_from_tag().
2. I thought the _lookup names were a little unclear (it's not obvious which direction they are going) so I renamed the other similar functions for consistency.
3. We had the function dcm_filehandle_read_filehandle_meta(), left over from when we renamed file as filehandle. I thought this was probably an error and it should be dcm_filehandle_read_file_meta().
4. We used _meta() in some places and _metadata() in others, which seemed inconsistent, so I switched to _metadata() everywhere.
5. I found an annoying bug in dataset_insert.

[hackermd]

@jcupitt One other thing that I just noted, which is not directly related to this PR: The _DcmVR enum not only contains individual VRs, but also options such as "OB or OW" (see here). Strictly speaking, those are not VRs and, in my opinion, therefore don't belong into the enum for the type definition. Can we handle those options differently?

[jcupitt]

I split the DcmVR enum into two, a public and a private one, and put the combination VRs into the private version.

There are two new things:

bool dcm_is_valid_vr_for_tag(DcmVR vr, uint32_t tag);
DcmVR dcm_vr_from_tag(uint32_t tag);

That can quickly check if a VR can be used for a known tag, and for implicit mode, a thing that can get the VR for a tag.

The new arrangement should also make it easy to add new VRs and VR combinations without breaking ABI.

[jcupitt]

Could someone merge this? I don't have permission. Once this merges, I have a few more small things that take it up to v0.2, the one that openslide builds and tests against.

[bgilbert]

Looks like tests haven't run. A re-push might fix it, but I think it can also happen if GitHub fails to create a merge commit to run tests against.

[jcupitt]

Oh, that's odd, it was showing all all complete for me yesterday. I'll push again.